A method for determining the center area of ​​a palm image

Abstract

The present invention provides a method for determining the center area of ​​a palm image, comprising thresholding a difference image of a palm image with a natural light-near infrared light combined light source and a palm image with a natural light light source, obtaining four finger root concave points between five fingers by using a clustering method in combination with the obtained corner points, locating the four finger root concave points, obtaining a palm area center line and a center point according to the positioning result, and finally determining the center area of ​​the palm image according to the finger root concave points and the palm area center point. The rectangular center area obtained by the determination method of the present invention is calculated adaptively by the size of the palm, has strong adaptability and robustness, and has a high accuracy rate of determination results.

Description

Technical Field

[0001] The invention belongs to the technical field of computer vision, and in particular relates to a method for determining the central area of ​​a palm image. Background Art

[0002] Determining the center area of ​​a palm image is the first step in palm print and palm vein image recognition, which determines the specific values ​​of subsequent palm print and palm vein features and has an important impact on recognition accuracy. Most existing palm localization algorithms use Harris corner points to extract finger key points, and then select the center area of ​​the palm based on empirical values. They have poor robustness and are greatly affected by lighting. The robustness and accuracy of positioning need to be improved and enhanced.

[0003] The following prior art documents are relevant to the present invention:

[0004] [1] Chen Diandian et al., Palm image ROI extraction method based on Harris corner detection and clustering algorithm. Optical Instruments, 2018, Vol. 40, No. 5: 27-34.

[0005] [2] Kuang Zhiqi et al., A new method for palm corner detection and palmprint region determination. Computers and Modernization, 2012, No. 205, No. 9, 2012: 123-126.

[0006] [3] Shang Li et al., A new palmprint ROI image positioning method. Laser & Infrared, 2012, Vol. 42, No. 7: 815-820.

[0007] [4]Koichi Ito et al. "Palm Region Extraction for Contactless PalmprintRecognition", International Conference on Biometrics, 2015: 334-340.

[0008] [5]Amira Oueslati et al, "An Efficient Palm Vein Region of InterestExtraction Method", International Conference on Advanced Technologies for Signal and Image Processing, 2020: IVP-54.

[0009] The ROI extraction method proposed in the literature [1] uses the Harris method to detect corner points, which will produce redundant corner points, resulting in poor robustness in the detection of finger pits. In addition, the image generated by the near-infrared sensor is not used. The thresholding of the palm image requires manual setting of the threshold, and the effect of binarizing the palm area is poor.

[0010] Reference [2] proposed to use the concave points between fingers to calculate the circumscribed circle of a triangle to obtain the center area of ​​the palm. The disadvantage is that the detection accuracy of the finger concave points has a great impact on the subsequent circular palm area. In addition, the Harris method is used in this paper to detect the finger concave points, and the results are seriously affected by lighting and palm posture.

[0011] In the literature [3], the height and width of the square area at the center of the palm are manually set, which has poor consistency for different palm center areas.

[0012] Reference [4] uses a fixed direction difference method to calculate the key points between fingers. It does not use a near-infrared sensor, so the palm binarization area obtained is poor, and this method is sensitive to the direction of the palm.

[0013] The palm area determination method proposed in the literature [5] requires the selection of control points in three cases to determine the concave points between fingers. It is highly complex and the width and height of the square area in the center of the palm need to be manually selected. There is room for improvement.

[0014] In summary, the robustness and accuracy of the palm center area determination method in the prior art need to be improved. Summary of the invention

[0015] In order to overcome the deficiencies of the prior art, the present invention proposes a new method for determining the center area of ​​the palm.

[0016] A method for determining the center area of ​​a palm image comprises the following steps:

[0017] (1) Subtracting the palm images collected under natural light and natural light-near infrared light combined light, and then preprocessing them to obtain a binary palm image;

[0018] (2) Obtain edge points of the binary palm image, extract corner points, and use clustering method to determine four finger root concave points in the binary palm image;

[0019] (3) locating the four finger root concave points, and using the located finger root concave points as key points to obtain the center line and center point of the palm area;

[0020] (4) Rotate the binary palm image so that the center line is horizontal, and determine the center area of ​​the palm image according to the coordinates of the key points and the center point in the rotated image.

[0021] In the above step (1), firstly, a palm image is collected by turning on the natural light source, and then a palm image is collected by turning on the natural light source and the near-infrared light source at the same time, and the difference image is obtained by subtracting the former from the latter. It is worth noting that the palm images collected under different light sources are images of the palm in the same posture.

[0022] Since the palm is close to the lens and the difference between the image collected by the background near-infrared light + natural light and the image collected using only natural light is small, the image collected by natural light + near-infrared light minus the image collected by natural light can obtain a binary image of the palm area more robustly and accurately.

[0023] This technical solution avoids the threshold selection process involved in directly using natural light images for binarization. Different thresholds need to be set for different lighting conditions, otherwise the extracted binary palm image will be inaccurate (i.e., over-segmentation or under-segmentation).

[0024] Obtaining the center point of the palm area in step (3) can also be performed after rotating the binary image in step (4).

[0025] Preferably, in step (1), the pretreatment comprises the following steps:

[0026] The difference image obtained after the difference is subjected to threshold processing, the connected domain with the largest area is retained and the holes in the largest connected domain are filled to obtain a binary palm image. As a further preferred embodiment, the threshold processing includes the following steps:

[0027] For all pixels in the difference image, compare them with a set threshold. If the pixel value is greater than or equal to the threshold, the value of the pixel is set to 1. If the pixel value is less than the threshold, the value of the pixel is set to 0.

[0028] As a further preferred embodiment, in step (2), obtaining edge points of the binary palm image and extracting corner points comprises the following steps:

[0029] Select any pixel with a value of 1 in the binary palm image as the current pixel. For the current pixel, if there is a pixel with a value of 0 among its 8 adjacent pixels, the current pixel is an edge point; otherwise, the current pixel is not an edge point.

[0030] Taking each edge point on the binary palm image as the center, if the ratio of the number of pixels with a value of 1 in the neighborhood of the set radius of the edge point to the number of all pixels in the neighborhood is greater than the set threshold, then the edge point is a corner point; otherwise, the edge point is not a corner point.

[0031] Preferably, in step (2), the four finger root concave points are determined by clustering method, which includes the following steps:

[0032] Store all corner points in the main set, select any corner point, store the corner point and the corner points whose distance to the corner point is less than the set threshold in the sub-set, and delete all corner points in the sub-set from the main set;

[0033] Repeat the above steps until the main set is empty, and obtain multiple subsets. Keep the four subsets with the top four corner points in the subsets. Calculate the mean of all corner points in each subset to obtain the class center of each subset, that is, the four root concave points.

[0034] The above steps are repeated as follows: continue to select the second corner point from the main set, store the second corner point and the corner points whose distance from the second corner point is less than the set threshold into the second subset, and delete all corner points in the second subset from the main set; repeat the operation until the main set is empty, and obtain multiple subsets.

[0035] Retaining the top four subsets in terms of the number of corner points in the subset should be understood as sorting all subsets in descending order according to the number of corner points, retaining the top four subsets, and performing subsequent operations.

[0036] Preferably, in step (3), the four finger root concave points are positioned by the following steps:

[0037] The sum of the distances between each finger base concave point and the other three finger base concave points is calculated respectively, among which the finger base concave point with the largest sum of distances between the other three finger base concave points is the thumb-index finger concave point; for the other three finger base concave points, according to the Euclidean distance from the thumb-index finger concave point from small to large, they are the index finger-middle finger concave point, the middle finger-ring finger concave point and the ring finger-little finger concave point.

[0038] As a further preferred embodiment, in step (3), the specific operation of taking the determined finger base concave point as the key point and obtaining the center line and center point of the palm area is as follows:

[0039] Taking the index finger-middle finger concave point, the middle finger-ring finger concave point and the ring finger-little finger concave point as key points, draw a straight line passing through the middle finger-ring finger concave point and perpendicular to the line connecting the index finger-middle finger concave point and the ring finger-little finger concave point to obtain the center line of the palm area;

[0040] Take the line segment where the center line intersects the binary palm image. The midpoint of the line segment is the center point of the palm area, and the pixel values ​​on the line segment are all 1.

[0041] In order to expand the central area of ​​the palm image, enrich the palm features extracted subsequently, and improve the palm recognition accuracy, optionally, in step (3), before obtaining the center line and center point of the palm area, the key points of the index finger-middle finger concave point and the ring finger-little finger concave point are adjusted as follows:

[0042] Connect the index finger-middle finger concave point and the middle finger-ring finger concave point to obtain a line segment, extend the line segment to the index finger direction by a set length, and use the end point of the line segment extension direction as a key point to replace the index finger-middle finger concave point; wherein the set length is 0.05 to 0.5 times the length of the line segment connecting the index finger-middle finger concave point and the middle finger-ring finger concave point;

[0043] A line segment is obtained by connecting the middle finger-ring finger concave point and the ring finger-little finger concave point, and the line segment is extended to a set length in the direction of the little finger, and the endpoint of the line segment extension direction is used as the key point of the ring finger-little finger concave point; wherein the set length is 0.05 to 0.5 times the length of the line segment connecting the middle finger-ring finger concave point and the ring finger-little finger concave point.

[0044] As a further preferred embodiment, the extension length of the line segment connecting the index finger-middle finger concave point and the middle finger-ring finger concave point is 0.2 times the length of the line segment;

[0045] The length of the extension of the line segment connecting the middle finger-ring finger concave point and the ring finger-little finger concave point is 0.2 times the length of the line segment.

[0046] As a further preference, step (4) is performed as follows:

[0047] Rotate the binary palm image to make the center line horizontal, calculate the coordinates of any point in the index finger-middle finger concave point and the ring finger-little finger concave point and the center point in the rotated image, calculate the difference in the number of rows and columns between the center point and the index finger-middle finger concave point or the ring finger-little finger concave point respectively, take twice the difference in the number of rows as the height and twice the difference in the number of columns as the width, draw a rectangle with the center point as the center and the width is horizontal, and the central area of ​​the binary palm image is obtained.

[0048] When rotating the binary palm image, first calculate the angle between the center line and the horizontal line. The index-middle finger concave point, the ring-little finger concave point and the center point are all two-dimensional coordinates. The first element in the coordinate represents the row number of the coordinate, and the second element represents the column number of the coordinate. The calculation formula for the angle θ between the center line of the palm area and the horizontal line is:

[0049]

[0050] Among them, p3(1) represents the row number of the index finger-middle finger pit or the ring finger-little finger pit, p3(2) represents the column number of the index finger-middle finger pit or the ring finger-little finger pit, p_c(1) represents the row number of the center point, and p_c(2) represents the column number of the center point.

[0051] The coordinates of the index-middle finger concave point, the ring-little finger concave point and the center point in the rotated image are calculated as follows:

[0052] Binarize any point p of the palm image before rotation. The point p is actually a two-dimensional coordinate. The first element represents the row number of point p, and the second element represents the column number of point p.

[0053] The rotation matrix is ​​calculated as follows:

[0054]

[0055] Calculate the center coordinates of the entire binary palm image, denoted as I_c;

[0056] The coordinates p′ after the rotation of point p are calculated according to the following formula:

[0057]

[0058] As a further preferred embodiment, after the central area of ​​the binarized palm image is determined, the natural light-near infrared light combination light source palm image and the binarized palm image are rotated by the same angle, and a rectangular image of corresponding size is intercepted at a position corresponding to the central area of ​​the binarized palm image in the rotated natural light-near infrared light combination light source palm image, so as to obtain the central area of ​​the palm image of the natural light-near infrared light combination light source palm image.

[0059] As a specific preference, a method for determining the central area of ​​a palm image comprises the following steps:

[0060] A1. Use visible light (natural light) to collect a palm image, denoted as A;

[0061] A2. Keep your palm still, turn on the visible light source and near-infrared light source at the same time, and collect another palm image, recorded as B;

[0062] A3. Subtract image A from image B, and record the result as image C (difference image);

[0063] A4. Thresholding image C to obtain a binary palm region image D;

[0064] A5. Keep the connected domain with the largest area in image D, and then fill the holes therein to obtain the binary palm image E;

[0065] A6. Calculate the edge points of the connected domain in image E, and denote the image composed of all edge points as image F;

[0066] A7. Extract corner points from all edge points to obtain the set of all corner points on image F. The set of corner points is denoted as S.

[0067] A8. Select the distance threshold T1, cluster all corner points in the set S according to T1, and only keep the four classes with the largest number of corner points (the top four);

[0068] A9. Calculate the class center of each of the four classes;

[0069] A10. Among these four cluster centers, find the cluster center with the largest sum of distances to the other three cluster centers. This cluster center is recorded as p1 (the thumb-index finger concave point).

[0070] A11. Find the cluster centers of the other three clusters closest to p1, and record them as p2 (index finger-middle finger concave point), p3 (middle finger-ring finger concave point), and p4 (ring finger-little finger concave point).

[0071] A12. Calculate a straight line passing through p3 and perpendicular to the line connecting p2 and p4, denoted as L (center line of the palm area);

[0072] A13. Calculate the midpoint of the pixel with a value of 1 on the straight line L on the image E, and record it as p_c (the center point of the palm area);

[0073] A14. Calculate the angle between the center line and the horizontal line

[0074] A15. Rotate image E counterclockwise by θ degrees, and record the rotated image as E';

[0075] A16. Change the pixel points of p2, p4, p_c in image E' to p2', p4', p_c';

[0076] A17. Calculate the absolute value of the difference in the number of rows and columns between any point in p2' and p4' and p_c', and record them as height and width respectively;

[0077] A18. With p_c' as the center, select a rectangular area on the image E' with 2*height and 2*width as the height and width of the rectangle, and record it as ROI (region of interest);

[0078] A19. Rotate image B counterclockwise by θ degrees, and extract the sub-image corresponding to the ROI on the rotated image B, that is, the central area of ​​the palm image, and output it.

[0079] Preferably, the image C is thresholded as follows:

[0080] Select threshold T2. For all pixels in image C, if the pixel value of a pixel is greater than or equal to T2, the value at the position corresponding to the pixel in D is 1; if the pixel value is less than T2, the value at the position corresponding to the pixel in D is 0.

[0081] Preferably, the binary palm image E is obtained by:

[0082] For any pixel with a value of 1 in image D, if the value of any pixel among the 8 adjacent pixels around it is 1, it is defined that the adjacent pixel is connected to the pixel, and all interconnected pixels form a connected domain. For all connected domains in image D, different numbers are used to mark the pixels in different connected domains. Then, the number of pixels in each connected domain is counted, and the values ​​of all pixels in the connected domain with the largest number of pixels are marked as 1, while the values ​​of all other pixels are marked as 0. The result is the binary palm image E.

[0083] Preferably, the edge points of the connected domain in the image E are calculated in the following manner:

[0084] Traverse all pixels in image E. If the value of a pixel is 1 and there is a pixel with a pixel value of 0 among its 8 neighboring pixels, define the pixel with a value of 1 as an edge point, and set the pixel value at the position of the pixel in image F to 1; otherwise, set the pixel value at the position of the pixel in image F to 0.

[0085] Preferably, the corner points are extracted from all edge points in the following manner to obtain a set of all corner points on the image F:

[0086] Define an empty set S;

[0087] For a pixel point (edge ​​point) with a value of 1 in the image F, a circular template with a radius of R is taken with the pixel point as the center, that is, the template value is 1 at the integer pixel position less than R from the pixel point, and the template pixel value is 0 at the integer pixel position greater than R;

[0088] Count the number of pixels with a value of 1 in the circular template, denoted as M;

[0089] Count the number of pixels of image E with a value of 1 in the circular template, M';

[0090] Set a ratio threshold T3 (between 0 and 1). If M' / M>T3, the pixel point (edge ​​point) is a corner point and is added to the set of corner points S. Otherwise, it is a non-corner point.

[0091] Preferably, all corner points in the corner point set S are clustered by the following steps:

[0092] First, create an empty set S1. Assume that the first corner point in S is a. Put a into S1 and delete a from S.

[0093] For each remaining corner point in the set S, calculate the Euclidean distance between the corner point and a. If the distance is less than the threshold T1, add the corner point to S1 and delete it from S.

[0094] Repeat the above two steps until the set S is empty;

[0095] Assume that we finally get N sets, which are denoted as S1, S2, …, SN. These N sets are also N subsets obtained by clustering the corner points, that is, N classes.

[0096] Preferably, the class center of each of the four classes is calculated by the following method:

[0097] For each class of corner points in the set, find the mean of all corner points, which is the class center of the class.

[0098] Compared with the prior art, the present invention has the following beneficial effects:

[0099] The method for determining the center area of ​​a palm image of the present invention uses a difference image of a palm image of a natural light-near infrared light combined light source and a palm image of a natural light light source for thresholding, uses a clustering method in combination with the obtained corner points to obtain four finger root concave points between five fingers, then locates the four finger root concave points, obtains the center line and center point of the palm area according to the positioning result, and finally determines the center area of ​​the palm image according to the finger root concave points and the palm area center point. The rectangular center area obtained by the determination method of the present invention is calculated adaptively by the size of the palm, has strong adaptability and robustness, and has a high accuracy rate of determination results. BRIEF DESCRIPTION OF THE DRAWINGS

[0100] Figure 1 A schematic flow chart of a method for determining the center area of ​​a palm image proposed in an embodiment of the present invention;

[0101] Figure 2 This is a schematic diagram of a palm image collected by using both natural light and near-infrared light in an embodiment of the present invention;

[0102] Figure 3 is a schematic diagram of a difference image after thresholding processing in an embodiment of the present invention;

[0103] Figure 4 Schematic diagram of a binarized palm image in an embodiment of the present invention;

[0104] Figure 5 Schematic diagram of a binary palm image edge image in an embodiment of the present invention;

[0105] Figure 6 A schematic diagram of the distribution of corner points in four subsets reserved in an embodiment of the present invention;

[0106] Figure 7 Schematic diagram of finger root concave points calculated in an embodiment of the present invention;

[0107] Figure 8 FIG. 1 is a schematic diagram of the index finger-middle finger concave point and the ring finger-little finger concave point after adjustment in the embodiment of the present invention (corresponding to the index finger point and the little finger point, respectively). The adjusted positions of the two are compared in FIG. Figure 7 The position in has changed;

[0108] Fig. 9 A schematic diagram of the center line of the palm area obtained in an embodiment of the present invention;

[0109] Fig.10 A schematic diagram of the index finger point, little finger point and palm area center point calculated by an embodiment of the present invention;

[0110] Fig.11 The determined position of the central area of ​​the binarized palm image in the embodiment of the present invention is shown in the area framed by the rectangle in the figure;

[0111] Fig.12 This is an image of the central area of ​​the palm captured from a palm image of a natural light-near infrared light combined light source in an embodiment of the present invention. DETAILED DESCRIPTION

[0112] The technical solution of the present invention is described clearly and completely below in conjunction with the embodiments of the present invention and the drawings of the specification. The embodiments described below are only a part of the present invention. All other embodiments based on the embodiments of the present invention and obtained without creative work are within the scope of protection of the present invention.

[0113] like Figure 1 As shown, a method for determining the central area of ​​a palm image comprises the following steps:

[0114] (1) Collect palm images under natural light and natural light-near infrared light respectively, make a difference between the two, and then perform thresholding to obtain a binary palm image.

[0115] Collect palm images under the combined light source of natural light and near-infrared light, such as Figure 2 As shown; when only the natural light source is turned on, a natural light palm image is taken, and the collection state of the palm is consistent with the collection state under the natural light-near infrared light combined light source.

[0116] Then adopt Figure 2Subtract the above natural light palm image from the palm image in (pixel value subtraction) to obtain a difference image;

[0117] According to the pixel value of the background area on the difference image, a threshold is selected to binarize the difference image so that the pixel value of the background area on the difference image is 0, and the pixel value of the palm area is 1. The binary image result is as follows: Figure 3 As shown;

[0118] right Figure 3 The binary image in the figure first finds the connected domain with the largest area, then sets all the pixel values ​​in other connected domains to 0, and only retains the connected domain with the largest area.

[0119] Fill the holes in the retained maximum connected domain to obtain Figure 4 Binarized palm image shown.

[0120] (2) Obtain the edge points of the binary image and extract the corner points, and use the clustering method to determine the four finger root concave points in the binary palm image. Figure 4 The edge points are obtained from the binary palm image in the figure. The specific method is to determine whether there is a pixel point with a value of 0 in the 8 neighboring pixels of each pixel point in the binary palm image. If there is, the pixel point is an edge point. The image composed of all edge points is as follows: Figure 5 As shown;

[0121] right Figure 5 For each edge point in , a circular template area is selected with the edge point as the center, and the number of edge points in the circular template area is calculated, and Figure 4 The number of pixels with a value of 1 in the circular template area is obtained by dividing the number of edge points by the number of pixels to obtain a ratio, and judging whether the ratio is greater than a set threshold. If the ratio is greater than the set threshold, the edge point is a corner point.

[0122] Store all the obtained corner points in the main set, set a distance threshold, randomly select a corner point, divide the corner point and all the corner points in the main set whose distance to the corner point is less than the threshold into the same sub-set, and delete all the corner points in the sub-set from the main set; repeat this cycle until the main set is empty, and obtain multiple sub-sets. Keep the top four sub-sets in terms of the number of pixels in the sub-sets; the corner points in these four sub-sets are Figure 4 The position distribution in Figure 6 shown.

[0123] Calculate the mean of all corner points in each of the four retained subsets, and the results are as follows Figure 7 As shown in the triangle point, Figure 7 The four triangular points are the four concave points between the five fingers;

[0124] (3) locating the four finger root concave points, and using the located finger root concave points as key points to obtain the center line of the palm area;

[0125] for Figure 7 The four finger base concave points found in , calculate the sum of the distances of each finger base concave point from the other three finger base concave points. The finger base concave point with the largest sum of distances from the other three finger base concave points is the concave point between the thumb and index finger; then calculate the Euclidean distances of the other three finger base concave points and the concave point between the thumb and index finger, among which the finger base concave point with the closest Euclidean distance to the concave point between the thumb and index finger is the concave point between the index finger and the middle finger, followed by the concave point between the middle finger and the ring finger, and the finger base concave point with the farthest Euclidean distance is the concave point between the ring finger and the little finger.

[0126] In order to expand the central area of ​​the palm image, enrich the palm features extracted subsequently, and improve the accuracy of palm recognition, the finger root concave points located above are adjusted:

[0127] For the concave point between the index finger and the middle finger, connect it with the concave point between the middle finger and the ring finger to obtain a line segment, and then extend the line segment along the direction of the index finger by a set length (0.2 times the length of the line segment). The endpoints of the extension direction of the line segment are as follows: Figure 8 As shown by the point where the triangle in the upper left corner is located, the index finger point is obtained to replace the concave point between the index finger and the middle finger;

[0128] For the concave point between the ring finger and the little finger, connect it with the concave point between the middle finger and the ring finger to get another line segment, and then extend the line segment to the little finger by a set length (0.2 times the length of the line segment). The endpoints of the extension direction of the line segment are as follows: Figure 8 As shown by the point where the triangle in the lower left corner is located, the little finger point is obtained to replace the concave point between the ring finger and the little finger;

[0129] The center line of the palm area is obtained by taking the index finger point, the concave point between the middle finger and the ring finger, and the little finger point as key points. The specific steps are as follows:

[0130] connect Figure 8 The point at the upper left corner (index finger point) and the point at the lower left corner (little finger point) are connected to obtain a line segment, and a straight line perpendicular to the line segment and passing through the concave point between the middle finger and the ring finger is drawn, such as Fig. 9 As shown by the straight line in, this straight line is the center line of the palm area;

[0131] (4) Rotate the binary palm image to make the center line of the palm area horizontal, and calculate the center point of the palm area; determine the center area of ​​the palm image according to the coordinates of the index finger point and the center point of the palm area in the rotated image.

[0132] Rotate the binary palm area image so that the center line of the palm area is in a horizontal position, rotate the palm image illuminated by a combination of near-infrared light and natural light by the same angle, and then find the line segment where the center line of the palm area intersects the palm area, that is, the line segment on the straight line where all pixel values ​​are 1;

[0133] Find the midpoint of the line segment, which is the center point of the selected palm area. Fig.10 The points indicated by the squares in the figure are shown.

[0134] calculate Figure 8 The difference in the number of rows and columns between the index finger point and the center point of the palm area are recorded as height and width respectively;

[0135] Taking the center point of the palm area as the center, and taking 2*height and 2*width as the height and width of the sub-image, a rectangular sub-image is intercepted on the rotated binary palm image, such as Fig.11 As shown in the rectangular area;

[0136] In the rotated natural light-infrared light combined light source palm image, capture Fig.11 The rectangular sub-image in is the sub-image at the corresponding position in the palm image, such as Fig.12 As shown in FIG. 4 , this is the center area of ​​the palm image determined by the method in this paper.

Claims

1. A method for determining the center area of ​​a palm image, characterized in that: The following steps are involved: (1) Subtracting the palm images collected under natural light and natural light-near infrared light combined light, and then preprocessing them to obtain a binary palm image; (2) Obtain edge points of the binary palm image, extract corner points, and use clustering method to determine four finger root concave points in the binary palm image; (3) locating the four finger root concave points, and using the located finger root concave points as key points to obtain the center line and center point of the palm area; (4) rotating the binary palm image so that the center line is horizontal, and determining the center area of ​​the palm image according to the coordinates of the key points and the center point in the rotated image; In step (2), four finger root concave points are determined by clustering method, including the following steps: Store all corner points in the main set, select any corner point, store the corner point and the corner points whose distance to the corner point is less than the set threshold in the sub-set, and delete all corner points in the sub-set from the main set; Repeat the above steps until the main set is empty, and obtain multiple subsets. Keep the four subsets with the top four corner points in the subsets. Calculate the mean of all corner points in each subset to obtain the class center of each subset, that is, the four root concave points. In step (3), the four finger root concave points are located by the following steps: The sum of the distances between each finger base concave point and the other three finger base concave points is calculated respectively, where the finger base concave point with the largest sum of distances between the other three finger base concave points is the thumb-index finger concave point; for the other three finger base concave points, the index finger-middle finger concave point, the middle finger-ring finger concave point and the ring finger-little finger concave point are ranked in descending order according to the Euclidean distance from the thumb-index finger concave point; In step (3), the specific operation of taking the determined finger base concave point as the key point and obtaining the center line and center point of the palm area is as follows: Taking the index finger-middle finger concave point, the middle finger-ring finger concave point and the ring finger-little finger concave point as key points, draw a straight line passing through the middle finger-ring finger concave point and perpendicular to the line connecting the index finger-middle finger concave point and the ring finger-little finger concave point to obtain the center line of the palm area; Take the line segment where the center line intersects the binary palm image, and the midpoint of the line segment is the center point of the palm area; Step (4) is performed as follows: Rotate the binary palm image to make the center line horizontal, calculate the coordinates of any point in the index finger-middle finger concave point and the ring finger-little finger concave point and the center point in the rotated image, calculate the difference in the number of rows and columns between the center point and the index finger-middle finger concave point or the ring finger-little finger concave point respectively, take twice the difference in the number of rows as the height and twice the difference in the number of columns as the width, draw a rectangle with the center point as the center and the width is horizontal, and the central area of ​​the binary palm image is obtained.

2. The method for determining the center area of ​​a palm image according to claim 1, characterized in that: In step (1), the pretreatment comprises the following steps: The difference image obtained after the difference is thresholded, the connected domain with the largest area is retained and the holes in the largest connected domain are filled to obtain a binary palm image.

3. The method for determining the center area of ​​a palm image according to claim 2, characterized in that: The thresholding process comprises the following steps: For all pixels in the difference image, compare them with a set threshold. If the pixel value of the pixel is greater than or equal to the threshold, the value of the pixel is set to 1. If the pixel value is less than the threshold, the value of the pixel is set to 0.

4. The method for determining the center area of ​​a palm image according to claim 3, characterized in that: In step (2), obtaining edge points of the binary palm image and extracting corner points includes the following steps: Select any pixel with a value of 1 in the binary palm image as the current pixel. For the current pixel, if there is a pixel with a value of 0 among its 8 adjacent pixels, the current pixel is an edge point. Taking each edge point on the binary palm image as the center, if the ratio of the number of pixels with a value of 1 in the neighborhood of the edge point with a set radius to the number of all pixels in the neighborhood is greater than the set threshold, then the edge point is a corner point.

5. The method for determining the center area of ​​a palm image according to claim 1, characterized in that: In step (3), before obtaining the center line and center point of the palm area, the key points of the index finger-middle finger concave point and the ring finger-little finger concave point are adjusted as follows: Connect the index finger-middle finger concave point and the middle finger-ring finger concave point to obtain a line segment, extend the line segment to a set length in the direction of the index finger, and use the endpoint of the line segment in the extension direction as a key point to replace the index finger-middle finger concave point; A line segment is obtained by connecting the middle finger-ring finger concave point and the ring finger-little finger concave point, and the line segment is extended to a set length in the direction of the little finger, and the endpoint of the extension direction of the line segment is used as the key point of the ring finger-little finger concave point.

6. The method for determining the center area of ​​a palm image according to claim 1, characterized in that: After the central area of ​​the binarized palm image is determined, the natural light-near infrared combined light source palm image and the binarized palm image are rotated by the same angle, and a rectangular image of corresponding size is intercepted at a position corresponding to the central area of ​​the binarized palm image in the rotated natural light-near infrared combined light source palm image to obtain the central area of ​​the palm image of the natural light-near infrared combined light source palm image.

Images